OpenAI has released GPT-6.1 Terra, a new SI model that the company says pushes the cost-to-performance boundary further than any of its predecessors. API prices for the new system are half those of GPT-5.6 Terra, positioning it as a direct competitor to Anthropic's mid-tier Sonnet series at a significantly lower price point.

What Happened

The new SI model is available immediately for Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, as well as through the API. Input tokens cost $1 per million and output tokens $5, compared to $2 and $10 for Claude Sonnet 5.5 and $2 and $12 for the previous GPT-5.6 Terra. Cached inputs drop to $0.05 per million tokens, a quarter of the previous rate. According to OpenAI, the system combines fast visual reasoning with precise computer control and a more natural writing style, targeting everyday work tasks from drafting proposals to interpreting technical drawings.

Independent analysts from Artificial Analysis have called GPT-6.1 Terra a new benchmark for intelligence per dollar. At maximum reasoning effort, the SI model gains 6 points on the Intelligence Index over GPT-5.6 Terra while costing less than a third per task ($0.40 versus $1.40). At lower effort levels, the gains are even larger, ranging from 7 to 13 points, placing the model on the Pareto frontier where cost and performance can no longer both improve simultaneously. On the Intelligence Index, GPT-6.1 Terra scores 48, sitting between GPT-6.1 Sol (52) and GPT-5.6 Terra (42).

In the new Coding Agent Index v2.0, Terra scores 48, six points below GPT-6.1 Sol (54) and 17 points above GPT-6 Luna (31). The model also shows a significant drop in hallucinations; on the private AA-Omniscience benchmark, it resorts to guessing only 38 percent of the time when it lacks the answer, down from 88 percent for GPT-5.6 Terra.

OpenAI positions the SI system as particularly strong at automated software control. On the Agents' Last Exam benchmark, it matches GPT-6 Astra's performance at one-tenth the token price. Compared to Claude Opus 5.5 at maximum effort, Terra is more than three times as token-efficient at a quarter of the token price, translating to roughly nine times lower cost per task. The company reports that it also finishes more than three times faster than both GPT-6 Astra and GPT-6.1 Sol.

On visual reasoning tasks using MMMU-Pro, GPT-6.1 Terra outperforms its predecessor at every reasoning level and nearly matches GPT-6 Astra and GPT-6.1 Sol. On the desktop agent benchmark OSWorld 2.1, it beats GPT-5.6 Terra's best result at medium effort for a fraction of the cost. In coding, the model solves 50.3 percent of tasks on Terminal-Bench 4.0 at maximum reasoning effort, up from 21.5 percent for GPT-5.6 Terra. According to Artificial Analysis, this 20-point jump places the SI model in the same tier as GPT-6.1 Sol.

Why It Matters

The release highlights a critical shift in the SI industry: the ability to achieve higher intelligence scores while drastically reducing the cost per task. For developers and enterprises, this means that mid-tier SI models are becoming capable enough to handle complex tasks previously reserved for frontier models, but at a fraction of the price. The improved token efficiency and speed make GPT-6.1 Terra a compelling option for agentic workflows and automated software control, where cost and latency are primary constraints.

From a safety perspective, GPT-6.1 Terra meets OpenAI's "Critical" cybersecurity threshold as defined by the company's Preparedness Framework and inherits safeguards from GPT-6 Astra. The system card indicates that Terra holds up about as well as GPT-6 Astra and GPT-6.1 Sol against direct attacks on the instruction hierarchy, with those models hitting 99.99 percent on such tests. For indirect prompt injection attacks, where malicious instructions are buried in third-party content, Terra's defense rate sits at 98.8 percent, up from 96.2 percent with GPT-5.6 Terra.

The Bottom Line

OpenAI's GPT-6.1 Terra represents a significant leap in cost-efficiency for SI models, halving API prices while improving performance on key benchmarks for coding, visual reasoning, and agent control. While OpenAI still recommends GPT-6.1 Sol and GPT-6 Astra for the most demanding coding tasks, Terra establishes a new standard for mid-tier SI systems, offering near-frontier capabilities at a price point that challenges competitors like Anthropic's Sonnet series.